In this paper we examine measurement error in the independent variables of multinomial response model. We show how not taking this error account biases the coefficient estimate asymptotically. We provide a bias-adjusted estimator, using the multinomial logit model as an example.
ASJC Scopus subject areas
- Economics and Econometrics